Historical Change in the Conceptualization of Gender in English Human-Generated Texts

Nela Riddle1, Alex Tran2, Amanda Diekman2


1 Department of Computer Science, Math
2 Department of Psychological and Brain Sciences

Introduction

Stereotypes about men and women have evolved over time, reflecting both real and perceived changes in gender roles. Prior research has shown that traits historically associated with men and women have shifted due to societal transformations (Eagly et al. 2020; Gustafsson Sendén, Bäck, and Lindqvist 2019; Diekman and Eagly 2000). While men and women are perceived to be becoming more similar in their traits, persistent gender gaps in agency and communion remain. However, these findings often rely on self-report data, which can introduce observer effects and inflate gender differences or similarities.

To overcome these limitations, we leverage large-scale, naturalistic human-generated text data as a measure of gender conceptualizations over time. Written language encodes societal beliefs, attitudes, and stereotypes about men and women, providing a rich source of insight into collective social representations (Bailey, Williams, and Cimpian 2022). By analyzing word embeddings, we can track the evolution of gender-trait associations from the 19th century to the present.

Guided by Social Roles Theory, we investigate how changes in gendered labor distributions correspond to shifts in trait perceptions. As gender segregation in social roles diminishes, we expect trait inferences about men and women to converge.

Purpose

Our study aims to expand upon previous findings by examining stereotype change on a larger scale using naturalistic data, providing insight into the long-term trajectory of gender conceptualizations.

Methods

Corpus EngAll COHA
Long Title Google Books Ngram Corpus, All English Corpus of Historical American English
Sources Books predominantly in the English language published in any country American text from the 1820s-2010s (fiction, magazine, newspaper, non-fiction, TV/movies)
Genre-balanced? No Yes
Total Words 850 billion 410 million
Decades 1800-1999 1820-2009

The Mean Average Correlation (MAC) was the main metric used, with the following process:

  • Compile word lists (trait, job, agentic, communal)

  • Compute MAC score between each word in list and “men,” “women” for a given decade

    • A higher score implies higher similarity between the group and term
  • Plot and compute Pearson correlation

  • Measure change in correlation over time; increase suggests groups have grown more similar with respect to a word list

Sample decade graphsSample decade graphs

Figure 1: Sample decade graphs

Results

Trends in role and trait terms

Figure 2: Trends in role and trait terms

Trends in communal and agentic traits

Figure 3: Trends in communal and agentic traits

Trends in magnitude, engall

Figure 4: Trends in magnitude, engall

Discussion

Over time, men and women have increasingly been conceptualized as occupying similar roles and possessing similar traits. This trend reflects a broader societal shift toward perceiving men and women as more alike in both agency and communion. Our findings indicate that the convergence in agency is larger than in communion, aligning with prior research (cf. Gustafsson Sendén et al., 2019). By utilizing a naturalistic and unobtrusive measure—human-generated texts—we provide an alternative to self-report methods, which may inflate gender similarities or differences due to observer effects.

Importantly, our study captures these shifts over a broad temporal scope, stretching back to the early 19th century. While historical shifts in language present challenges, such as semantic drift (e.g., the word “gay” shifting in meaning over time), our approach mitigates this concern by analyzing word proximity to men and women at the same time points. This ensures that changes in word meaning do not compromise our measure of gender similarity in conceptualization.

References

Bailey, April H, Adina Williams, and Andrei Cimpian. 2022. “Based on Billions of Words on the Internet, People= Men.” Science Advances 8 (13): eabm2463.
Diekman, Amanda B, and Alice H Eagly. 2000. “Stereotypes as Dynamic Constructs: Women and Men of the Past, Present, and Future.” Personality and Social Psychology Bulletin 26 (10): 1171–88.
Eagly, Alice H., Christine Nater, David I. Miller, Michèle Kaufmann, and Sabine Sczesny. 2020. “Gender Stereotypes Have Changed: A Cross-Temporal Meta-Analysis of u.s. Public Opinion Polls from 1946 to 2018.” American Psychologist 75 (3): 301–15. https://doi.org/10.1037/amp0000494.
Gustafsson Sendén, Marie, Emma Bäck, and Anna Lindqvist. 2019. “Introducing a Gender-Neutral Pronoun in a Natural Gender Language: The Influence of Time on Attitudes and Behavior.” Frontiers in Psychology 10: 37. https://doi.org/10.3389/fpsyg.2019.00037.